Thursday, 23 May 2013

NBN: Not just "there's good, cheap, fast: choose any two"

Paul Wallbank, a knowledgable and respected technology journalist, wrote recently on the NBN using the "pick any two" meme. There are exceptions to this Ironclad Rule of Project Management and my contention is the NBN is one of them. Paul wrote:
One of technology’s truisms is there are three factors to almost anything – speed of delivery, reliability and affordability (or cheap, good, quickly) – and you can choose any two. If you choose cheap then you have to be prepared to sacrifice either speed or reliability.

I throughly agree with Paul's following conclusion:
However, just fixating on the advertised speeds of both network proposals misses essential aspects of the debate and its uses in a connected economy and online society.

We need to choose carefully and make some informed decisions that accurately reflect the needs of todays and tomorrow’s internet users.
This Ironclad Rule, or perhaps only Rule of Thumb, presumes the design, technology, process and people/capability of competing approaches are identical or comparable. It's like comparing sports teams, some are better than others, either overall or in particular areas.

Background: These aren't Apples and Apples projects

The differences in the Coalition and ALP proposals for the NBN significantly affect their projects:
  • The majority services of the two plans, GPON/FTTP and DSL/FTTN, have radically different cost and performance structures:
    • Fibre has effectively unlimited upgrade capacity.
      • Allowing much more finely differentiated plans and reducing consumer surplus and markedly increasing profits over the assumed "single plan" of the FTTN.
      • Higher line speeds will radically increase consumption of high-end users, markedly increasing total usage/revenues & profits, driving down usage charges and stimulating download across all subscribers and encouraging up-selling.
    • While DSL/FTTN over copper is economically limited in Australia to 25-50Mbps.
    • Fibre components & transceivers are like computers and following Moore's Law. Both increasing rapidly in capability/speed/power efficiency, but also declining in price.
    • Signal processing hardware necessary for DSL/FTTN to increase speed is complex, expensive and very limited production volumes. Like "supercomputers" of old, it's a dead-end technology that can only get less cost-competitive.
    • The installation costs of Fibre are well-understood, though like any large construction project, this does not remove Project Risk. Whilst NBN Co has allocated a 10% contingency, unanticipated problems, especially with non-compliant or defaulting contractors already seen in the Northern Territory, with the major rollout phase yet to commence and ramp-up, this must be assessed as a "High Risk".
      • The budget & schedule impact on the ~6-year $10-15 billion mass fibre roll-out has the potential to force a 50%-75% over-run in cost. The project phase is highly parallel and additional resources/teams can be trained and deployed to keep the schedule.
    • The maintenance costs of the copper network to the much more stringent VDSL2 levels are uncertain and unknown.
      • The Coalition is completely silent on who will wear these extra charges.
      • Will NBN Co directly assume all maintenance costs for the copper Customer Access Network
    • The uncertainties in the DSL/FTTN rollout over copper are legion and severe:
      • Not only are there multiple "drop-deads" in this FTTN project, the Coalition has not discussed their risk, the downside cost, mitigation strategies, alternatives or 
      • Will the Telstra shareholders even agree to the change in plan? It's not a management, or even Board, decision to make.
        • Given Telstra has a history of driving very hard bargains and using dedicating large legal teams to major contracts, will the deal forced on the Coalition allow a DSL/FTTN to ever be profitable, let alone break-even?
      • Will VDSL2 at 30Mhz work on decades old 0.4mm copper at 800m even at 25Mbps?
      • What if the distance rule for nodes turns out to be 300m requiring 400,000 nodes, not 68,000 are required?
        • This scenario will also force a much higher rollout costs on NBN Co for whom the Coalition has said it will cap funding at around $30 billion.
      • What if 200-port nodes cost closer to $150,000 than $30,000 to install, test and cut-over?
      • Since 2005, Telstra FTTN plans have allocated a major fraction of the project to reconfiguration and remediation of the Copper Network. What if the copper network has degraded further than anyone expects?
        • When will the Coalition know this?
        • When will they pull they plug on the project if costs are astronomical?
      • By allowing competition and cherry-picking, what if mobile and fixed-line services from Telstra, Optus and other players undercut the NBN Co network in the low-cost, high-density urban areas, leaving NBN Co with a very low connection rate and the most costly customers?
        • NBN Co must plan to provide universal coverage.
        • We know from the both the 1994/6 Cable TV rollout and withholding ADSL2  that Telstra and Optus are highly aggressive players and will force competitors to make investments that become uneconomic and waste billions.
        • With the Coalitions' funding cap to NBN Co of around $30 billion, will they even be able to complete the FTTN before running out of cash?
      • If Telstra & Optus run NBN Co out of cash and steal all their subscribers, what happens to the public investment then? 
    • Demand growth and implied usage increase driven by new Applications or increase in uptake of services like IPTV.
      • NBN Co uses a conservative estimate of 30%/year compound annual growth rate (CAGR), or 32 month doubling period for demand.
      • More realistic estimates are 50% and 63% (ABS and CISCO VNI report).
    • Wholesale price model:
      • NBN Co will reduce wholesale download charges 19%/year for 30% growth, a 5.3%/year effective increase.
      • The Coalition models include a 1% increase above inflation in APU (Average Revenue Per User).  We don't know if this will be an instruction to NBN Co, a outcome of capped demand due to lower speeds or an estimate of traffic.
    • The Coalition has not supplied a public sector economic "Cost Benefit Analysis" (CBA) for DSL/FTTN, such as Henry Ergas' 2009 submission to Senate Broadband inquiry. After very stridently demanding CBA's from the ALP, this is not just ironic but a incontrovertible double-standard.
      • The Coalition in their Broadband plan implicitly recognise that they are building the DSL/FTTN to throw away. Otherwise, why do they mandate NBN Co must both provide on-demand Fibre upgrades and 
      • Whatever the bluster & noise the Coalition puts out, a copper Customer Access Network has a limited economic life in Australia, especially for broadband carriage.
      • In the Coalition modelling provided, two critical data points are deliberately omitted:
        • What is the economic break-even for the DSL/FTTN network?
        • When do they expect the copper Customer Access Network, for phone and broadband, will be shutdown?
        • No, this won't be "up to NBN". It's a central assumption in their modelling.
    In summary, Fibre leads to a virtuous circle of increasing demand, lower prices and higher usages, leading to higher profits: this is the economics that O.T.C. traded on skilfully for over 20 years.

    In contrast, an FTTN leads to a "death spiral" of high costs, under-cutting, users fleeing to substitutes in droves and plummeting revenues. Exactly what IBM experienced leading up to its near demise in 1991/2.

    Why the Ironclad Rule doesn't apply to NBN and FTTN vs FTTP 

    The differences in Capital and Operational Expenditure of a full Fibre network and a majority DSL/FTTN network are quite extreme, especially with market growth potential, availability of substitutes, nature and number of competitors, production economics and project risk factors and potential downsides.

    But the underlying project differences are the transmission technologies:

    • Right now, commodity Optical Fibre transmitters are 1Gbps and soon 10Gbps, with 40Gbps and 100Gbps in production. This isn't speculation or "shown in the Lab", it's current commercial volume production.
      • Whilst the proponents of digital-on-copper spruik Gigabit rates as possible, and 100Mbps as almost commercially available('vectoring'), these are untried technologies implemented with very complex, expensive signal-processing and based on the presumption of high-quality customer copper lines.
      •  The correct capability comparison of Fibre and DSL/copper is 1000:1. 100Gbps (high-end Fibre) versus 100Mbps DSL + 'vectoring'. Costs are not that dissimilar at the high-end.
      • The real cost comparison for Fibre and Copper is "same-same" or 50-100Mbps.
        • A VDSL2 line-card may be $75-$100 per end per port, but
        • Fibre transceivers for 100Mbps are in the $20-$50 range.
        • It's no contest on the technology at 100Mbps on cost, maintenance or link quality. Which is why the Coalition has to adamantly and vehemently insist that "nobody could need more than the speed of a donkey".
      • Cost & link quality comparisons at higher speeds, even the commodity line rate of 1Gbps are so skewed in favour of Fibre they aren't even worth contemplating.
    • Economic distances for low Australian housing densities are 400m-800m for digital transmission over copper. The copper is mostly thin, 0.4mm, radically limiting maximum raw rates. The only way a DSL/FTTN can be competitive is to use existing infrastructure and incur near zero additional labour costs. There are multiple additional operations that must be performed in rolling out a DSL/FTTN:
      • building nodes by the ten-thousand, with council approvals, electricity connection, dual fibre uplinks, cable trunks to pillars and multiple terminations millions of subscriber lines for both phone and broadband.
      • reconfiguring and rehabilitation, remediation or replacement of copper.
      • per-service line-testing, fault diagnosis and rectification.
        • Probably multiple times per line as more lines are cutover
      • Service identification, cut-over and consequential fault reporting and correction.
        • Because the exchange line is cut, the subscriber service must be made work, there is no fallback.
          •  This only happens in GPON/FTTP with "pull-throughs", where the lead-in is in such bad condition that the existing copper is used to pull-through the new fibre. But if this is done, a replacement copper lead-in could often be run as well.
        • With 9 million lines and 4-5 million active services, the fault rate on conversion/cut-over will be high: 10-20% at least.
        • Fixing these "own-goal" faults will be costly and time-consuming, particularly because the customer service will be inoperative and service fines will apply.
        • Expect the ACCC to insert itself in this area. 
    With all technologies, things change. A hundred years ago, horses were the mainstay for local transport and steam trains for long-distance and bulk haulage.

    In this milieu, Cobb and Co were the Telstra of their day. They were the largest, most profitable, vertically integrated service supplier around. Nobody could compete with them on reach, price, service availability or speed & punctuality. These were extraordinary businessmen running a phenomenal and efficient business.

    Within 25 years their business had collapsed.
    Like Microsoft post 2005, they had failed to move into new areas of demand and not just adopt, but fully embrace the new technologies.

    Sticking with a copper network, with limited capability, high install and maintenance costs and acute economic susceptibility to substitution from not one, but two, aggressive mobile (4G) network operators is as economically doomed as Cobb and Co not embracing

    Think of the radical changes in computing technology we've seen in just the last 10 years. This is the timeframe that we must judge a DSL/FTTN within - that is its longest economic life:

    • laptops, netbooks and games consoles (Wii, XBox, Playstation, ... [not  my area of expertise])
    • smart phones, tablets, ebook readers and "mobile devices"
    • 300Mbps Wifi
    • Ubiquitous USB and bluetooth connectivity and swarms of devices
    • GPS, compasses, inertial sensors in mobile devices
    • Cloud computing, commodity virtual machines and pay-per-hour compute services, such as Amazon AWS used by Netflix.
    • 3G and 4G data services, although shared medium, speeds comparable with cable TV
    • PVR's, multi-Terrabyte Hard disks and 100GB Solid State Disks (flash)
    • digital cameras (with video), iPods, $5-10 multi-GB flash memory cards/USB sticks
    • Digital TV (DVB-T) and satellite, 16:9 HD broadcast (1920x1080p),
    • Plasma then LCD large format screens, "Home Theatres",
    • Social Media: Facebook, Twitter, picture & video services

    Individually, they are radical advances, together they are game changers, like the many variants of the motor-vehicle, in all aspects of our lives, businesses and economy.

    There is no sign that the rate of I.T. Innovation will slow, either for hardware, software or Internet services.

    In this environment, stating "This limited bandwidth is all anyone will ever need" is an extraordinary statement, as bizarre as the misquoted US Commissioner of Patents in 1898,  "Everything that can be invented has been invented".

    Carl Sagan posited "Extraordinary claims require extraordinary evidence".
    That test needs to be applied to the Coalition's seemingly blind-faith assertion that a DSL/FTTN will provide everything all broadband consumers could ever want.

    Instead of meekly accepting this outrageous claim, the electorate needs to be asking the Coalition for their extraordinary evidence to back their DSL/FTTN policy.

    No comments:

    Post a Comment

    Note: only a member of this blog may post a comment.